InferMap
Map ambiguous CSV or API column headers to your target schema before writing migrations or import code.
Overview
InferMap is a MCP server for the Build phase that maps messy table columns to your known schema using multi-scorer inference.
What is this MCP server?
- Maps messy columns to a known schema with seven scorers and domain dictionaries
- Advertised F1 0.84 with zero configuration in catalog copy
- MCP 0.5.0 via Railway streamable-http or stdio through PyPI infermap and uvx
- Complements dedup pipelines—clarify fields before GoldenMatch or GoldenPipe
- Agent-callable alternative to manual column guesswork in import scripts
- Seven scorers plus domain dictionaries per description
- Advertised F1 0.84 zero-config mapping
- MCP server version 0.5.0
What problem does it solve?
Unknown column names block imports because you cannot trust which field is email, SKU, or amount.
Who is it for?
Indie builders ingesting third-party CSV or JSON feeds who have a target schema but not a data engineer.
Skip if: Fully standardized APIs with OpenAPI you already codegen, or unstructured document extraction problems.
What do I get? / Deliverables
Your agent proposes schema-aligned column maps so transforms and migrations target the right fields.
- Proposed column-to-field mapping with scorer-backed confidence
- Faster import script generation from aligned schema
- Reduced wrong-key deduplication when paired with Golden tools
Recommended MCP Servers
Journey fit
Schema inference belongs in build when you connect external feeds to your models and need column alignment first. InferMap solves header-to-field mapping at integration boundaries, not frontend UI or launch distribution.
How it compares
Column-to-schema inference MCP, not duplicate detection (GoldenMatch) or full pipeline orchestration (GoldenPipe).
Common Questions / FAQ
Who is InferMap for?
Solo developers using AI agents to wire imports when source headers do not match internal database field names.
When should I use InferMap?
Early in build integrations before ETL code, deduplication, or load tests on ambiguous exports.
How do I add InferMap to my agent?
Configure the InferMap Railway MCP endpoint or local stdio with uvx and the infermap PyPI package in your agent MCP file.